Sharing LR Bypass Gain Capacitor possible? (CE Amplifier)

[domingo]

I need to fit two Common Emitter Amps (stereo) into a very small box, to reduce space usage I thought of sharing a 1000u bypass cap that gives extra gain (C_3) between both channels.

Image #1 shows L R amps as they currently are (and work).

Image #2 shows what I have in mind. I know this might be silly because AC current from one channel might return to the other channel, but since I don't understand how the amp works exactly and the cap goes to ground I thought it might work.

Would it work?

Thanks,
Domingo

 

[moamps]

IMO, this will probably work considering that both amplifiers are very close to each other and the resulting -3dB corner is below 10Hz .. But I would have some remarks on the values of the other parts ....

 

[Bo Deadly]

With separate 51R it would probably work fine. Without the 51R you might get a tiny amount of crosstalk depending in the layout. Ideally the ground of the 1000uF should be right next to both emitter resistors or at least equidistant. But with the 51R, the "ideal" is not nearly as important.

 

[domingo]

Thank you for your reply @moamps. The diagram only shows the first stage for simplicity, the full diagram of each amp goes attached here and you are welcome to comment on it. I tested many configurations using a CE calc until I found this one to be the most satisfactory (to feed an opamp headphone amp, source is a 2.2k impedance electret).

Back to the "shared cap" experiment, would it still work if I share the cap among three preamps instead of two? And can I also share the 51ohm resistor (R_7), so instead of using three 51ohm resistors I use only one; meaning to share the whole bypass line (resistor+cap) among 3 preamps?

 

[domingo]

With separate 51R it would probably work fine. Without the 51R you might get a tiny amount of crosstalk depending in the layout. Ideally the ground of the 1000uF should be right next to both emitter resistors or at least equidistant. But the 51R, the "ideal" is not nearly as important.

Oh you can kind of answered my previous new question on the 51ohm resistor. Better to keep them separate apparently... They are actually 3 preamps, difficult to keep them very close but equidistant yes, with cables of 5cms for + and - from preamp(s) to cap, would that be ok?

What do you mean by "But the 51R, the "ideal" is not nearly as important."? That the value of the 51R is not that important? :s

 

[Bo Deadly]

Oh you can kind of answered my previous new question on the 51ohm resistor. Better to keep them separate apparently... They are actually 3 preamps, difficult to keep them very close but equidistant yes, with cables of 5cms for + and - from preamp(s) to cap, would that be ok?

What do you mean by "But the 51R, the "ideal" is not nearly as important."? That the value of the 51R is not that important? :s

I mis-typed. I meant to type "But with the 51R, the" ideal of being close / equidistant is not as important.

The 51R is doing a bunch of stuff so if you trust whoever designed this thing, I would just leave it. It's controlling the gain but it's also providing feeback. A smaller resistor gives you higher gain (potentially much higher) but more distortion (potentially much more). The 51R is also affecting the input impedance which is approximately (51R + emitter resistance) * transistor gain which is probably something like (51+25)*200 = 15K. But that is paralleled with the two bias resistors, which means the input impedance is something like 2K. The electret is seeing something like 2.2K + 2K.

Anyway, if you shared the 51R, all of this stuff would break and it would be a complete disaster. You would get major crosstalk at least.

The electrolytic has some non-insignificant impedance, particularly at low frequencies. If you have a larger cap you would lower that impedance at lower frequencies. And note that the voltage of that cap can be low like 10V or maybe even 6.3V depending on the supply voltage. So even a 4700uF doesn't have to be huge.

 

[domingo]

When you talk about equidistant is it mostly referring to the distance between 51R and Cap, or Cap and GND? Or both are equally important? I'm adding a SPST switch between Cap and GND common to all Preamps to turn ON/OFF the extra gain of the bypass line and I'm wondering if the length of all cables should be the same. Two preamps should have symetric gain (stereo), the third preamp can be slightly different because it belongs to a summed different third mono line.

I didn't know that 51R affects input impedance, but didn't understand how the two paralleled bias resistors reduce impedance from 15K to 2K. Can I assume that if I reduce the actual source resistor (2.2K) to 220ohm overall input impedance will come down to something around 2.2K? I need to roughly match the electret's 2.2K output impedance.

 

[DKatz]

Switch disconnecting common cap would greatly increase crosstalk. Cap ground should be close to inputs ground IMO.
You don't need to match mic impedance, higher is OK.

 

[Bo Deadly]

When you talk about equidistant is it mostly referring to the distance between 51R and Cap, or Cap and GND? Or both are equally important? I'm adding a SPST switch between Cap and GND common to all Preamps to turn ON/OFF the extra gain of the bypass line and I'm wondering if the length of all cables should be the same. Two preamps should have symetric gain (stereo), the third preamp can be slightly different because it belongs to a summed different third mono line.

It doesn't matter in this case. I was being pedantic. You can have 5cm of wire to the cap to GND (or what we prefer to call "0V" for a signal ground) no problem. That is all very low impedance so any inductance or resistance in the wire is going to be completely dwarfed by the 51R at the other end so noise and stability is not an issue.

Now if you put the 51R at the cap and ran a long wire to the transistor, that would be bad because now your 5cm wire has an impedance of 51R. Still not high but high enough to make me wince and mutter to myself. So if you want the third channel to have a different value, change it at the emitter of the transistor and not at the switch.

I didn't know that 51R affects input impedance, but didn't understand how the two paralleled bias resistors reduce impedance from 15K to 2K. Can I assume that if I reduce the actual source resistor (2.2K) to 220ohm overall input impedance will come down to something around 2.2K? I need to roughly match the electret's 2.2K output impedance.

You cannot change that 2.2K because it will affect the bias of that transistor. Descrete circuits like this are delicately balanced. You don't need to match the impedance of the electret. It's customary to go higher. And as it is it looks like it's at least 4K. I would just keep the values as is. You can switch in the cap to change gain. Thats fine. But I would just keep the values of things as they are (although you can use a bigger cap).

This all assumes the person who designed this thing knew what they were doing of course.

 

[Bo Deadly]

Switch disconnecting common cap would greatly increase crosstalk.

True. Good point.

However, if 3 separate traces (or wires in this case) were used to connect to 3 poles of the switch so that the 3 circuits were only connected together when the common capacitor is in the circuit, then it would work without added crosstalk.

 

[domingo]

if 3 separate traces (or wires in this case) were used to connect to 3 poles of the switch so that the 3 circuits were only connected together when the common capacitor is in the circuit, then it would work without added crosstalk.

In the attachment goes what I have in mind... I interpret from the answers that this would add crosstalk, but why would that be the case? If the 0V to all PAs is common in any case...

PD: A SPDT is drawn because LibrePCB comes without SPST switches.

PD2: Should the ground of the "Gain Cap" better go to the emitter resistor 0V or the electret GND?

 

[Bo Deadly]

I interpret from the answers that this would add crosstalk, but why would that be the case? If the 0V to all PAs is common in any case...

But not when the cap is switched out. When the cap is switched out you have signal between the amps through the 51R that are all connected together. You would get major crosstalk. With the cap switched in, the point where they're all connected together is effectively grounded at AC and so no cross talk.

The solution is to use a 3 pole switch with each 51R connected to a separate pole so that when they're switched out, they're not connected to each other. When the cap is switched in, they all connect to the + of the cap and grounded at AC.

Ideally you could use some shielded 3+ conductor cable that runs out to the switch and uses the shield as a return to the cap +. Then you can put the cap near the emitters of the transistors. But again with the 51R in series, the "ideal" is not nearly as important. The signal wires are all very low impedance when the cap is in which is swamped by the 51R or very high impedance when the cap is switched out. So shielding is actually not vitally important either.

 

[domingo]

When the cap is switched out you have signal between the amps through the 51R that are all connected together. You would get major crosstalk.

I see it now!

I don't have 3 pole switches but I could use a 2 pole for the stereo line and single pole for the mono right? This would also allow me to switch on the extra gain to the mono or the stereo line independently, a plus

In the picture what I interpret is a correct solution.

 

[JohnRoberts]

The schematic looks like it says 1uF.

Using only one cap for multiple channels will likely degrade LF separation but may be inconsequential if low enough frequency.

Voltage on the ground side of that cap will look like a - input, so where it is connected in the layout could impact different channels differently.

JR

 

[domingo]

Using only one cap for multiple channels will likely degrade LF separation but may be inconsequential if low enough frequency.

Could you please elaborate on the reason why sharing the cap will degrade LF separation? Is it because of the cap no having enough capacitance for all three meeting currents, or else?

 

[JohnRoberts]

Could you please elaborate on the reason why sharing the cap will degrade LF separation? Is it because of the cap no having enough capacitance for all three meeting currents, or else?

at low enough frequency that cap will look like an open circuit, so all the 51 ohm resistors will appear connected to each other and feed signal directly into each other.

Another issue is that for signal integrity the - lead of that cap wants to be at the same node as the input ground... impossible if one caps is shared between multiple channels.

JR

 

[domingo]

at low enough frequency that cap will look like an open circuit, so all the 51 ohm resistors will appear connected to each other and feed signal directly into each other.

Thanks for the clarification. Would that fall into the audible range in this case? With the 1000U cap and 1K collector resistor here I assume (but might be wrong) that the -3dB corner in theory is 0.16Hz. Would you say that anything below that crosstalk starts?

The issue of signal integrity wouldn't be solved if the (-) of the cap is hooked directly to the (-) of each electret instead of the (-) of the emitter resistor? :s

 

[Bo Deadly]

Another issue is that for signal integrity the - lead of that cap wants to be at the same node as the input ground... impossible if one caps is shared between multiple channels.

Would the 51R in series not dwarf any effects of wire / trace impedance between the - lead of the cap and the input?

 

[JohnRoberts]

This is not about trace impedance...

with a common emitter gain stage the base of the transistor looks like the + input of a differential amp and the emitter a low impedance - input. This may be hard to visualize but ground is a concept not a voltage. There will be tiny voltage potential differences between the ground trace node at different points. The voltage gain for these tiny voltage errors will be -1000/51, you do the math. In an ideal world everything connected to the ground trace will be at 0V but we do not operate in an ideal world.

I have written expensively about differential stages, not as much about c-e gain stages, but they share some math.

JR

 

[Bo Deadly]

This is not about trace impedance...

with a common emitter gain stage the base of the transistor looks like the + input of a differential amp and the emitter a low impedance - input. This may be hard to visualize but ground is a concept not a voltage.

Ok. Nice visualization. I follow.

There will be tiny voltage potential differences between the ground trace node at different points.

Right. The tiny differences are due to resistance (and less so inductance) of the trace. Being a cap to 0V, I don't suspect parasitic capacitance matters much.

The voltage gain for these tiny voltage errors will be -1000/51, you do the math. In an ideal world everything connected to the ground trace will be at 0V but we do not operate in an ideal world.

But -1K/51R is a LOT less than what it would be without the 51R. And we have to include emitter resistance in this case so it's actually more like 25+51 and that is comparable to the parallel 100R which is going to diminish the effects of trace impedance further.

With a high open loop gain op amp circuit that has a 10R gain resistor and no emitter resistance making 60dB, then putting your cap 5cm away would be a bad design decision. But this circuit is none of those things.